Dr G. K. T. Holmes, Department of Gastroenterology, Derbyshire Royal Infirmary, London Road, Derby DE1 2QY, UK. E-mail: email@example.com
Background In view of the high diagnostic accuracy of immunoglobulin-A-tissue transglutaminase antibodies for detecting coeliac disease, we have explored whether a small bowel biopsy is always required to establish the diagnosis.
Aim To define the transglutaminase antibody level giving a positive predictive value for coeliac disease of 100% and to subsequently assess the proportion of new diagnoses of coeliac disease having such a result.
Methods The Celikey kit (Phadia GmbH, Frieburg, Germany) was used to measure transglutaminase antibody levels.
Results All patients with transglutaminase antibody levels >30 U/mL, i.e. 10 × upper limit of normal in 2002/2003 had characteristic small bowel mucosal lesions. In a subsequent audit, 58% of 112 new diagnoses of coeliac disease in 2004/2005 had levels above this cut-off value.
Conclusions We have shown that a transglutaminase antibody level can be defined which gives a positive predictive value of 100% for coeliac disease. From published data, these observations can be extended to most second-generation transglutaminase antibody kits. Our data provide further evidence that diagnostic guidelines could be modified so that small bowel biopsy is no longer regarded as mandatory in patients with such high transglutaminase antibody levels. This will avoid an invasive procedure and lead to a more rapid diagnosis and earlier treatment for over half of the new patients with coeliac disease.
First-generation tests1 for serum IgA-tissue transglutaminase antibodies (TGA) showed inferior performance to endomysial antibody in routine use, with a low positive predictive value (PPV) for the detection of coeliac disease (CD).2 However, second-generation assays for the detection of TGA, using human purified or human recombinant tissue transglutaminase-2 (tTG) as antigen, have high sensitivity and specificity. In a prospective study of serum samples received from 1554 subjects, we reported an overall PPV of 85%.3 Systematic reviews have concluded that TGA has high diagnostic performance in children and adults for identifying those with CD.4, 5 The role of TGA in the investigation of CD has recently been reviewed.6
Data from the United Kingdom National External Quality Assessment Scheme and from comparative studies show that TGA kits from different manufacturers vary in their analytical and diagnostic performance7 and laboratories should ensure that they are using a thoroughly validated method.6 A comparative study of 10 second-generation kits noted that the number of calibrators varied from 1 to 6, which influenced the accuracy at higher levels. Two kits that employ a single standard showed poor linearity (R2 = 0.603 and 0.844) compared to kits using three or more standards (R2 = 0.991–0.996).8 Manufacturers of both of these kits now provide five calibrators. Where multiple calibrators are used, results are directly estimated from the calibration curve and non-linearity at higher levels is therefore not an issue; with a single calibrator, the method assumes a linear relationship between absorbance and TGA units. For the kit showing R2 of 0.603, linearity was maintained to only four times the upper limit of normal (ULN). Such procedures are therefore inherently inaccurate at higher levels and do not provide accurate quantitative results unless laboratories observe strict procedures for the dilution of samples above defined levels. Manufacturer’s diagnostic cut-offs should always be confirmed, preferably using the range of samples routinely received by the laboratory. Although absolute values are not interchangeable between different manufacturers, correlation between results from samples within the measuring ranges of the kits was generally acceptable.8 A second large comparative study also noted a high degree of agreement between kits for samples from patients with CD.9 To compare the data from different manufacturers, in the absence of an international standard, they expressed results as the ratio between the value obtained and the decision threshold based on receiver operating characteristic curves (i.e. as multiples of the ULN).
With the advent of an accurate, quantitative serological test, the requirement for small bowel biopsy to establish the diagnosis of CD in every case has been questioned.10–12 If biopsy could be avoided, then some patients would be spared an uncomfortable procedure, and treatment with a gluten-free diet (GFD) to alleviate symptoms prescribed at an earlier period.
The aims of this study were to assess: (i) the TGA level at which the PPV for CD was 100%, (ii) the proportion of new diagnoses of CD in our centre for the years 2004 and 2005 with TGA above that level and (iii) for these patients, the delay in diagnosis and treatment due to the present requirement for small bowel biopsy.
Patients and Methods
To assess the positive predictive value of IgA-tissue transglutaminase antibody at different cut-off levels. Adults (age >15 years) with TGA >10.0 U/mL measured in the period between April 2002 and December 2003 and with a small bowel biopsy were included. Duplicate TGA results on individual patients were excluded on the following basis: for new diagnoses of CD in the review period, the result immediately prior to biopsy was used. For subjects with a histological diagnosis of CD established prior to the review period, the first result reported in the review period was included. For patients with a normal biopsy, the highest result was included.
Distribution of IgA-tissue transglutaminase antibody levels in new diagnoses of coeliac disease. Transglutaminase antibody results measured prior to, or less than 21 days after, a small bowel biopsy from adults (age >15 years) with a new diagnosis of CD in 2004 and 2005 were included. Patients on gluten restriction or those with IgA deficiency were excluded.
Transglutaminase antibody was measured with human recombinant tTG as antigen (Celikey; Phadia GmbH, Freiburg, Germany). Results were calculated from the calibrators provided with the kit (maximum calibrator was 100 U/mL). The zero calibrator was assigned a value of 0.1 U/mL in order to calculate results from a logarithmic calibration curve as described by the manufacturer. We have previously shown at a cut-off of 3 U/mL (the ULN) that the sensitivity and specificity for the diagnosis of CD are 92% and 98.9% respectively.3 Routine inter-assay imprecision at 0.2 and 26 U/mL was 24% and 7.6% respectively. The laboratory participates in the United Kingdom External Quality Assessment Scheme for coeliac serology.
Interval between serology and small bowel biopsy
For those patients with a new histological diagnosis of CD in 2004/2005 and with TGA at a level consistent with a PPV for CD of 100%, the interval (days) between the date of serology and the date of biopsy was noted. When there was more than one TGA result greater than this cut-off, then the date of the earliest result was used.
Diagnosis of coeliac disease
Diagnosis of CD was based on jejunal biopsy evidence of villous atrophy, increased intra-epithelial lymphocytes and crypt hyperplasia. Biopsies were evaluated using a modification of the standardized report scheme based on the Marsh classification.13
This study was approved by the Derbyshire Local Research Ethics Committee.
Assessment of the positive predictive value of IgA-tissue transglutaminase antibody at different cut-off levels
Small bowel biopsy results were available on 148 patients with TGA >10 U/mL between April 2002 and December 2003. Results from 2 were excluded because of the long interval (17 and 24 months) between a normal biopsy and subsequently abnormal serology. The distribution of the TGA results on the remaining 146 patients (male: n = 43, age 22–89 years, median 48 years; female: n = 103, age 19–84 years, median 42 years) are shown in Figure 1. Of these patients, 139 had CD and the Marsh classification of the small bowel biopsies was: type 2, n = 10; type 3A, n = 16; type 3B, n = 60; type 3C, n = 53. Seven patients had TGA results between 10 and 30 U/mL (range: 11.1–21.8 U/mL) and normal biopsies. In five of these endomysial antibody was positive (TGA range: 12.8–21.8 U/mL). Table 1 shows the cumulative data at each of the TGA cut-offs expressed as PPVs for the diagnosis of CD.
Table 1. The positive predictive value (PPV) for different cut-offs of IgA-tissue transglutaminase antibody (TGA) levels in 146 patients
Not coeliac disease
* TGA is expressed as U/mL and as multiples of the upper limit of normal (ULN).
Distribution of IgA-tissue transglutaminase antibody levels in new diagnoses of coeliac disease
One hundred and twelve patients with a new diagnosis of CD between January 2004 and December 2005 fulfilled the criteria for inclusion in the study. The Marsh classification of the small bowel biopsies was: Type 2, n = 7; Type 3A, n = 14; Type 3B, n = 46; Type 3C, n = 45. TGA was >30 U/mL (>10 × ULN) in 65 (58%).
Interval between serology and small bowel biopsy for patients with a new diagnosis of coeliac disease and transglutaminase antibody level ≥30 U/mL
The mean interval between serology and biopsy was 108 days (range 6–319 days) for the 65 patients whose TGA was >30 U/mL (>10 × ULN). Table 2 shows the distribution of the interval for these patients. The interval was >90 days for 50% of patients.
Table 2. The interval (days) between serology and biopsy for patients whose transglutaminase antibody level was >30 U/mL
Interval (days) between serology and biopsy
No. patients (%)
Small intestinal histology is considered the ‘gold standard’ for the diagnosis of CD but because of the advent of reliable screening tests, this is being questioned with some advocating that, with a consistent clinical picture and positive serological tests, a biopsy is not mandatory.10–12 In a paediatric population,12 48 of 49 children with a TGA level >5 × ULN had positive biopsy results, although others have shown that the TGA test used has poor linearity.8 We have shown in adults that a TGA level >30 U/mL (>10 × ULN) using the Celikey test kit is absolutely predictive for CD when judged against characteristic small bowel mucosal appearance. The highest level observed in those patients without biopsy evidence of CD was 21.8 U/mL (7.3 × ULN). Five of the seven patients with normal biopsies had positives endomysial antibodies as well as increased TGA; it is possible that some of these will develop CD in future.14 The ability to define a level that gives a PPV of 100% for the diagnosis of CD, adds further evidence that small bowel biopsy can be avoided in some patients. In order to explore the implications for routine clinical practice, we reviewed TGA results in 112 newly diagnosed patients with CD in 2004 and 2005, to assess the proportion with a level giving a PPV of 100%. 58% had such a result and could therefore, have been spared a small bowel biopsy. For these patients, the delay between the serological result and confirmation of the diagnosis by biopsy varied between 6 and 319 days, with 48% and 12% waiting more than 90 and 180 days respectively. These worrying figures occurred because of cumulative delays in making a referral to hospital, waiting for endoscopy and for the report of small bowel biopsies. The major factor in the delay was the time between the general practitioner receiving the serological result and the gastroenterologist receiving the referral for endoscopy. The subsequent wait for endoscopy was usually 2 weeks and not more than 4 weeks. Thus, almost two-thirds of endoscopies presently undertaken in this situation to diagnose CD would no longer be required and treatment with a GFD brought forward by an average of about three months.
Serological tests have high diagnostic accuracy in samples requested from primary care3 and patients who are informed of positive results are often reluctant not to exclude dietary gluten immediately especially when they have troublesome symptoms. General practitioners at this stage also feel that they should be intervening and may be under pressure to prescribe a GFD. Gluten exclusion can often make subsequent interpretation of small bowel morphology difficult as improvement towards normal begins to occur. Some patients who have improved clinically are reluctant to reintroduce gluten prior to endoscopy to provoke mucosal change for fear of developing symptoms, or even to have the procedure carried out at all. There is already evidence that the diagnosis of CD is increasingly based on serological evidence alone. For the patients served by our laboratory and gastroenterological services (population 500 000), of those with positive serology between the years 1996 and 2000, only 11% were not biopsied. Between 2001 and 2005, 34% were not referred for biopsy. This difference can be attributed specifically to the acceptance of serology as diagnostic, because the number of patients refusing biopsy or where it was contraindicated because of co-morbidity can be assumed to be similar in the two quinquennia. Throughout this decade, positive serology reports returned to requesters, have included the same advice to refer patients to the coeliac clinic for further assessment.
Our data confirm that a biopsy is not required when TGA is >10 × ULN. However, a biopsy is necessary when TGA is increased but is below this cut-off, in order to exclude the diagnosis in those with false-positive serology although a proportion of these are likely to have potential CD and will develop overt CD in due course.14 As about 8% of patients with CD have negative serology at diagnosis, a biopsy is also indicated in those with negative serology but a clinical picture consistent with CD.3, 15 Even if biopsies are obtained, these can pose difficulties. In one study, 11% of biopsies were of such poor quality that they could not be interpreted.16 In addition, biopsies showing mild histological lesions are difficult to interpret and although quantitative morphometry and immunohistochemistry can help to indicate CD, most pathology laboratories will not have these techniques available. Until now, advocates of serology have interpreted TGA results as either positive or negative. When the cut-off is set at the ULN, the PPV for a positive test is 75–85%, hence the requirement to proceed to small bowel biopsy as some of those with positive tests may have normal mucosa or minimal change. The value of this study is that we have shown that a TGA level can be defined above which all patients show the mucosal lesions characteristic of CD. We have used a cut-off of 10 × ULN; however, as the highest TGA result observed in a patient with a normal biopsy was 7.3 × ULN, we could have used a lower cut-off e.g. 8 × ULN.
An important consideration is whether the data we have obtained using one manufacturer’s kits is applicable to results obtained by TGA assays developed by other suppliers. Different manufacturers use a variety of forms of tTG-2 as antigen for detecting the serum antibody. Additionally, there is no numerical agreement between manufacturers because no international reference material is yet available to provide uniformity of calibration and reference ranges. However, recent comparative studies.8, 9 show that, with few exceptions, there is acceptable agreement between second-generation kits. A comparison of the diagnostic accuracy of ten second-generation assays8 showed that with 2 exceptions, there was excellent linear correlation between kits from different manufacturers. Comparisons based on multiples of an appropriate ULN should therefore be valid.
The second comparison9 assessed the overall agreement between 11 different kits and looked particularly at the problem of false-positive TGA results in the patients with cirrhosis. Like the previous study, they showed good agreement between kits for samples from patients with CD. In eight of nine second-generation kits, none of the false-positive results in cirrhotics were more than three times the ULN. The authors considered that for the ninth second-generation kit, the ULN had been set too low by the manufacturer; the two patients (out of 54) with TGA results >3 × ULN were both below that cut-off when the ULN was corrected based on receiver operating curve data. Two third-generation kits, using gliadin peptides combined with tTG as antigen, both gave higher false-positive results in small numbers of samples from cirrhotics. False-positive results may rarely occur in other diseases but only at low titre.
These two studies are important because they demonstrate good agreement between most second-generation kits based on multiples of the ULN and show that false positives in liver disease are well below our suggested cut-off (10 × ULN). Different manufacturer’s kits vary in the measuring range, e.g., the Phadia Celikey kit measures up to 100 U/mL which is 33 × ULN, whereas the Euroimmun kit measures up to 220 U/mL which is 11 times their proposed cut-off of 20 U/mL. It is important that laboratories become familiar with available tests and base their selection on objective criteria,6 so that there is confidence in test performance. The practice of reporting results as ratios of a single standard, i.e. providing a semi-quantitative result, should be abandoned. Quantitative results, based on a calibration curve, with additional sample dilution when results are above the highest calibrator, should become the accepted laboratory practice.
In conclusion, we have shown that the PPV for CD is 100% when, TGA is >10 × ULN. On the basis of literature evidence, we believe that similar results will also be found with other second-generation kits. If these data can be confirmed from other centres or lower cut-offs defined which give a PPV of 100% for CD, then we suggest that it is time to consider modifying national and international guidelines so that small bowel biopsy is no longer regarded as mandatory for the diagnosis of CD in patients with such high TGA levels. Removing the requirement for small bowel biopsy will avoid an invasive procedure and lead to more rapid diagnosis with the potential for earlier treatment.
Declaration of personal and funding interests: Part of the expenses paid for P. G. Hill to attend the 12th International Symposium on Coeliac Disease, New York, November 2006, were met by Phadia.